City three-dimensional terrain rapid modeling method and system based on AI point cloud semantic segmentation

Abstract

The application relates to the technical field of computer 3D modeling, and discloses a city three-dimensional terrain rapid modeling method and system based on AI point cloud semantic segmentation. According to city terrain semantic point cloud fusion preprocessing data set and in cooperation with a point cloud three-dimensional modeling platform, high-quality three-dimensional modeling of feature objects in a target city terrain is carried out, and precise modeling under a city terrain limited point cloud data state is realized based on semantic segmentation and point cloud fusion. Through the point cloud three-dimensional modeling platform and screenshot software, city terrain three-dimensional model image information is accurately collected, and intelligent identification of city terrain three-dimensional model modeling defects is carried out in combination with AI algorithms and city terrain feature object standard modeling defect image data. Based on city terrain three-dimensional model data and in cooperation with the point cloud three-dimensional modeling platform, dynamic and efficient repair and reconstruction processing of city terrain three-dimensional model modeling defects is carried out, and precise modeling and repair and reconstruction under a city terrain limited point cloud data state are realized based on model repair.

Description

Technical Field

[0001] This invention relates to the technical field of computer 3D modeling, specifically to a method and system for rapid urban 3D terrain modeling based on AI point cloud semantic segmentation. Background Technology

[0002] The technical background of point cloud semantic segmentation modeling stems from the rapid development of 3D perception technology and the urgent needs of application scenarios. The widespread use of devices such as LiDAR and depth cameras has reduced the cost of acquiring point cloud data and improved its accuracy. This type of data presents the geometric information of the object surface through a set of 3D coordinate points and has the advantage of being unaffected by changes in lighting. However, the inherent sparsity, irregularity, and noise interference of point clouds pose a severe challenge to traditional 2D image processing methods, requiring specialized algorithms to achieve point-by-point semantic annotation. The use of multilayer perceptrons and symmetric functions to extract global features lays the foundation for deep learning to process unordered point sets. With the increasing demands for scene understanding accuracy in fields such as autonomous driving and industrial inspection, semantic segmentation technology needs to identify targets such as roads, obstacles, and equipment components in complex environments in real time, driving researchers to continuously optimize feature extraction, context modeling, and multimodal fusion capabilities. Current technological trends focus on lightweight real-time models, improved anti-interference robustness, and cross-modal collaborative learning to meet the application needs of scenarios such as the Industrial Internet of Things. Currently, autonomous driving requires accurate object recognition and modeling of urban 3D terrain environmental features in the vehicle driving environment. However, during the autonomous vehicle's operation, it is impossible to accurately collect urban terrain point cloud data from all directions, resulting in poor quality of urban 3D terrain feature modeling based on point cloud data. Existing autonomous driving urban 3D terrain modeling cannot achieve accurate identification and high-quality modeling of urban terrain feature objects under limited point cloud data conditions.

[0003] Chinese invention patent application CN116958420A discloses a high-precision modeling method for 3D facial features of a digital human teacher. This method involves using 66 RGB cameras distributed to collect omnidirectional images of the teacher's face, employing an array of cameras to capture images from multiple perspectives, and obtaining scattered point data. The collected scattered point data is then denoised and filtered to improve the quality of the generated point cloud data. A 3DMM model is constructed based on the point cloud data from multiple perspectives. PointCNN is used to learn the features and semantic information of the point cloud, segmenting it into different facial regions. For each perspective of the facial point cloud, its feature representation is combined with the 3DMM model for 3D facial reconstruction. Compared to traditional 3DMM models, the PointCNN model can extract richer point cloud features and more accurately capture shape details. However, the above technical solutions cannot accurately identify the teacher object and perform precise 3D facial modeling with limited point cloud data. Summary of the Invention

[0004] To address the existing limitations of accurate identification and high-quality modeling of urban 3D terrain features in autonomous driving systems with limited point cloud data, this project aims to achieve the following: real-time acquisition of urban terrain point cloud data; autonomous identification of standard point cloud data for target urban terrain features; accurate generation of semantic point cloud data for urban terrain; scientific construction of fused semantic point cloud data for urban terrain; scientific generation of preprocessed fused semantic point cloud data for urban terrain; dynamic construction of 3D urban terrain models; intelligent analysis of modeling defects in 3D urban terrain models; accurate reconstruction of 3D urban terrain models; and improved quality of rapid 3D urban terrain modeling.

[0005] This invention is achieved through the following technical solution: a rapid urban 3D terrain modeling method based on AI point cloud semantic segmentation, the method comprising the following steps:

[0006] The process involves collecting urban terrain point cloud data, searching a standard point cloud database of feature objects in the target urban terrain to obtain a standard point cloud database of feature objects, identifying the attributes of feature objects in the urban terrain point cloud to obtain semantic point cloud data of the urban terrain, fusing the semantic point cloud of the urban terrain with the standard point cloud of the urban terrain feature objects to obtain fused semantic point cloud data of the urban terrain, performing data denoising preprocessing on the fused semantic point cloud data of the urban terrain to obtain fused preprocessed semantic point cloud data of the urban terrain, modeling the 3D model of feature objects in the target urban terrain to obtain 3D model data of the urban terrain, collecting image data of the 3D model of the urban terrain, identifying modeling defects in the 3D model of the urban terrain to obtain modeling defect identification information of the 3D model of the urban terrain, directly outputting the 3D model of the urban terrain when there are no defects, and repairing and reconstructing the modeling defects of the 3D model of the urban terrain when defects exist to obtain 3D reconstructed model data of the urban terrain and outputting it.

[0007] Preferably, the following steps are taken: First, collect urban terrain point cloud data. Then, search and process the standard point cloud database of feature objects in the target urban terrain to obtain a standard point cloud database of target urban terrain feature objects. Next, identify the attributes of the feature objects in the urban terrain point cloud to obtain urban terrain semantic point cloud data. Finally, fuse the semantic point cloud of the urban terrain with the standard point cloud of the urban terrain feature objects to obtain fused semantic point cloud data. Finally, perform data noise reduction preprocessing on the fused semantic point cloud data of the urban terrain to obtain preprocessed semantic point cloud fusion data of the urban terrain.

[0008] By collecting and processing point cloud data of feature objects in the surrounding terrain environment of an autonomous vehicle using an onboard LiDAR, a set of urban terrain point cloud data is obtained. ,in Indicates the number of collections Urban terrain point cloud data; the urban terrain point cloud data includes the location information and color information of the outline surface of the target urban terrain feature object;

[0009] Based on the urban terrain point cloud data set and the standard point cloud database matrix of urban terrain feature objects, a search process is performed on the standard point cloud database of feature objects in the target urban terrain to obtain a standard point cloud database set of target urban terrain feature objects; based on the urban terrain point cloud data set and the standard point cloud database set of target urban terrain feature objects, feature object attribute identification processing is performed on the urban terrain point cloud to obtain an urban terrain semantic point cloud data set.

[0010] Based on the urban terrain semantic point cloud data set and the target urban terrain feature object standard point cloud database set, the semantic point cloud of urban terrain and the standard point cloud of urban terrain feature object are fused to obtain an urban terrain semantic point cloud fused data set; based on the urban terrain semantic point cloud fused data set, the urban terrain semantic point cloud fused data is denoised and preprocessed to obtain an urban terrain semantic point cloud fused preprocessed data set.

[0011] Preferably, the steps for searching the standard point cloud database of feature objects in the target city terrain based on the urban terrain point cloud dataset and the standard point cloud database matrix of urban terrain feature objects to obtain the target city terrain feature object standard point cloud database set are as follows:

[0012] Establish a standard point cloud database matrix for urban terrain feature objects. ,in Indicates the first A standard point cloud database of urban terrain feature objects corresponding to various types of urban terrain feature objects, wherein the types of urban terrain feature objects include different types of ground, different types of buildings, different types of vehicles, different types of transportation equipment, different types of vegetation, and different types of people; the standard point cloud database of urban terrain feature objects represents a collection of point cloud data of the complete feature object spatial contour surface set for different types of urban terrain feature objects. ,in The standard point cloud database representing the urban terrain feature objects The Middle The standard point cloud data of urban terrain feature objects includes the location information, color information, and type attribute information of the outline surface of the urban terrain feature objects; wherein the type attribute information of the urban terrain feature objects includes any one or more of the following: name information, specification information, type information, gender information, and material information of the urban terrain feature objects.

[0013] The urban terrain point cloud data set was collected using a point cloud 3D modeling platform. All the urban terrain point cloud data mentioned in the text The spatial point cloud contour features of the urban terrain feature objects and the standard point cloud database matrix of the urban terrain feature objects constituted The internal standard point cloud database of urban terrain feature objects Standard point cloud data of all urban terrain feature objects mentioned above The spatial point cloud contour features of the standard urban terrain feature objects are matched to search for matching points cloud data sets. The standard point cloud database of urban terrain feature objects is used to match the spatial point cloud contour features of urban terrain feature objects. And through data identification, a standard point cloud database set of target city terrain feature objects is generated. , ,in and They represent the first species and first A standard point cloud database of target city terrain feature objects corresponding to various types of city terrain feature objects. ,in and These respectively represent the standard point cloud database of the target city terrain feature objects. The Middle The and the first Standard point cloud data of the terrain features of each target city; ,in and These respectively represent the standard point cloud database of the target city terrain feature objects. The Middle The and the first Standard point cloud data of target city terrain feature objects, wherein the point cloud 3D modeling platform includes any one of Geomagic Studio, CloudCompare, and Open3D; the standard point cloud data of target city terrain feature objects includes the position information, color information, and type attribute information of the outline surface of the target city terrain feature objects.

[0014] Based on the point cloud 3D modeling platform, the urban terrain point cloud data set will be used. The urban terrain point cloud data described in The target city terrain feature object standard point cloud database set is used for matching the spatial point cloud contour features of urban terrain feature objects. The target city terrain feature object standard point cloud database described in the article to The corresponding urban terrain feature object type text information is mapped to the urban terrain point cloud data that matches the spatial point cloud contour features of the target urban terrain feature object in the standard point cloud database. And through data identification, a set of urban terrain semantic point cloud data is generated. ,in Indicates the first The city terrain semantic point cloud data includes the location and color information of the outline surface of the target city terrain feature object, as well as the type attribute information of the target city terrain feature object.

[0015] Preferably, the steps for fusing the semantic point cloud of urban terrain with the standard point cloud database of the target urban terrain feature objects based on the urban terrain semantic point cloud data set to obtain a fused urban terrain semantic point cloud data set; and the steps for performing noise reduction preprocessing on the fused urban terrain semantic point cloud data set to obtain a preprocessed urban terrain semantic point cloud data set are as follows:

[0016] Based on the point cloud 3D modeling platform, it will be combined with the urban terrain point cloud data set. The urban terrain point cloud data described in The target city terrain feature object standard point cloud database set is used for matching the spatial point cloud contour features of urban terrain feature objects. The internal target city terrain feature object standard point cloud database to The standard point cloud data of the target city terrain feature object and the set of semantic point cloud data of the city terrain are described in the text. The urban terrain semantic point cloud data described in Point cloud data is combined by matching the spatial point cloud contour features of urban terrain features, and a semantic point cloud fusion dataset of urban terrain is generated. , > ,in Indicates the first The city terrain semantic point cloud fusion data includes the position and color information of the fused outline surface of the target city terrain feature object, the type attribute information of the target city terrain feature object, and the position and color information of the target city terrain feature object after fusion.

[0017] The urban terrain semantic point cloud fusion data set is obtained from the point cloud 3D modeling platform. Located in the middle, composed of all the aforementioned urban terrain semantic point cloud fusion data The urban terrain semantic point cloud fusion data within the spatial point cloud features of the urban terrain feature objects constituted. Perform data filtering and preprocessing, and then select the data located in the fusion data of all the urban terrain semantic point clouds. The urban terrain semantic point cloud fusion data constitutes the spatial point cloud feature contour surface of the urban terrain feature object. Data identification is performed to generate a preprocessed dataset for urban terrain semantic point cloud fusion. , ≤ ,in Indicates the first The preprocessed data for the fusion of semantic point clouds of urban terrain includes the position and color information of the fused contour surface of the target urban terrain feature object, and the type attribute information of the target urban terrain feature object.

[0018] Preferably, the following steps are taken: First, model the 3D model of the feature objects in the target city terrain to obtain 3D city terrain model data. Second, collect 3D city terrain model image data, identify and process modeling defects in the 3D city terrain model to obtain 3D city terrain modeling defect identification information. Third, when no defects are found, directly output the 3D city terrain model.

[0019] Based on the urban terrain semantic point cloud fusion preprocessing data set, the three-dimensional modeling process of feature objects in the target urban terrain is carried out to obtain urban terrain three-dimensional model data.

[0020] Based on the urban terrain 3D model data, the appearance image of the urban terrain 3D model is acquired and processed to obtain an urban terrain 3D model image data set. Based on the urban terrain 3D model image data set and the urban terrain feature object standard modeling defect image data matrix, the modeling defect identification processing of the urban terrain 3D model is performed to obtain urban terrain 3D model modeling defect identification information. The urban terrain 3D model modeling defect identification information includes no defects and defects. When there are no defects, the urban terrain 3D model is directly output.

[0021] Preferably, the steps for performing 3D modeling of feature objects in the target city terrain based on the urban terrain semantic point cloud fusion preprocessing data set to obtain urban terrain 3D model data are as follows:

[0022] Obtain the urban terrain semantic point cloud fusion preprocessing data set. ;

[0023] Based on the urban terrain semantic point cloud fusion preprocessing data set, a point cloud 3D modeling platform was used. The urban terrain semantic point cloud fusion preprocessing data described in the article The corresponding target city terrain feature object integrates the position and color information of the outline surface and the type attribute information of the target city terrain feature object to perform three-dimensional model classification and modeling processing of the feature object in the target city terrain, and generates urban terrain three-dimensional model data, which represents the surface outline three-dimensional model information of the target city terrain feature object.

[0024] Preferably, based on the urban terrain 3D model data, the appearance image of the urban terrain 3D model is acquired and processed to obtain an urban terrain 3D model image data set. Based on the urban terrain 3D model image data set and the standard modeling defect image data matrix of urban terrain feature objects, modeling defect identification processing of the urban terrain 3D model is performed to obtain urban terrain 3D model modeling defect identification information. The urban terrain 3D model modeling defect identification information includes whether there are no defects or whether there are defects. When there are no defects, the operation steps for directly outputting the urban terrain 3D model are as follows:

[0025] The urban terrain 3D model data is input into a point cloud 3D modeling platform and run, and displayed in all directions on a screen. Simultaneously, screenshot software is used to capture images of the displayed urban terrain 3D model, generating a set of urban terrain 3D model image data. ,in Indicates the number of collections Three-dimensional model image data of urban terrain;

[0026] Establish a standard modeling defect image data matrix for urban terrain feature objects ,in Indicates the first The image data represents standard modeling defects of urban terrain feature objects; the standard modeling defects of urban terrain feature objects represent standard urban terrain 3D model appearance image information set for the defect types of urban terrain feature object point cloud 3D modeling; wherein the defect types of urban terrain feature object point cloud 3D modeling include gaps on the model surface, uneven model surface, irregular bumps on the model surface, burrs or isolated floating patches on the model surface, and sharp edges on the model surface.

[0027] The set of urban terrain 3D model image data The urban terrain 3D model image data described in the document The image data matrix of defects in the standard modeling of the urban terrain feature objects The urban terrain feature object standard modeling defect image data described in the article Image feature matching is performed, and based on the image feature matching results, defect identification information for the 3D urban terrain model is generated. The specific steps for generating the defect identification information for the 3D urban terrain model are as follows:

[0028] Step 2231, Input Image Preprocessing: Preprocess the urban terrain 3D model image data set. The urban terrain 3D model image data described in the document and the standard modeling defect image data matrix of the urban terrain feature object The urban terrain feature object standard modeling defect image data described in the article The image data is preprocessed by scaling the image to a fixed size and normalizing the image.

[0029] Step 2232, Convolutional Layer Processing: Process the preprocessed urban terrain 3D model image data and the standard modeling defect image data of the urban terrain feature objects Convolution is used for processing; the convolution formula is: ,in and These represent the first convolution adjustment factor and the second convolution adjustment factor, respectively. Represents a constant. This represents the input variable of the convolution formula, namely the preprocessed 3D urban terrain model image data. and the standard modeling defect image data of the urban terrain feature objects ; express The convolution objective function, The base is The sum of the exponents is The exponential function; Represents the convolution objective function and exponential function In the interval The convolution value, Indicates the input variables of the convolution formula The derivative; after extraction and convolution, a new 3D model image data of the urban terrain is formed. and the standard modeling defect image data of the urban terrain feature objects ;

[0030] Step 2233: Activation function layer processing: Process the 3D urban terrain model image data output by the convolutional layer. and the standard modeling defect image data of the urban terrain feature objects Nonlinear transformations are performed using the ReLU activation function, the formula for which is: ,in Represents the three-dimensional model image data of the urban terrain and the standard modeling defect image data of the urban terrain feature objects The convolution value; Indicates the value Random numbers within the interval;

[0031] Step 2234, Pooling layer processing: Processing the linearly transformed urban terrain 3D model image data... and the standard modeling defect image data of the urban terrain feature objects Perform a downsampling operation;

[0032] Step 2235, Fully Connected Layer Processing: Process the downsampled urban terrain 3D model image data and the standard modeling defect image data of the urban terrain feature objects Flattened into vectors, these vectors can be connected to fully connected layers for classification or regression.

[0033] Step 2236, Output Results: Output the three-dimensional urban terrain model image data to the fully connected layer. and the standard modeling defect image data of the urban terrain feature objects Image matching processing is performed based on the urban terrain 3D model image data. and the standard modeling defect image data of the urban terrain feature objects Image feature matching results are used to generate a 3D urban terrain model and identify defects in the modeling process.

[0034] when and If image feature matching is successful, it indicates that there are defects in the modeling of urban terrain features and the model needs to be repaired and rebuilt. In this case, the defect identification information of the three-dimensional urban terrain model is output as "defect exists".

[0035] when and If no matching of image features is found, it indicates that there are no defects in the modeling of urban terrain features. In this case, the defect identification information of the urban terrain 3D model is output as "no defects". At this time, the urban terrain 3D model data is directly input into the point cloud 3D modeling platform and run in conjunction with the display screen to output the urban terrain 3D model.

[0036] Preferably, when defects exist, the steps for repairing and reconstructing the modeling defects of the urban terrain 3D model to obtain and output the urban terrain 3D reconstruction model data are as follows:

[0037] When the defect identification information of the urban terrain 3D model is that there is a defect, the urban terrain 3D model data in the urban terrain 3D model data is used to perform defect detection, repair and reconstruction work on the model outline surface modeling of the urban terrain 3D model through the point cloud 3D modeling platform, and urban terrain 3D reconstruction model data is generated.

[0038] The urban terrain 3D reconstruction model data is input into the point cloud 3D modeling platform and run, and the urban terrain 3D model is output in conjunction with the display screen.

[0039] A rapid urban 3D terrain modeling system based on AI point cloud semantic segmentation is used to implement the rapid urban 3D terrain modeling method based on AI point cloud semantic segmentation. The system includes an urban terrain point cloud processing module, an urban terrain modeling and defect analysis module, and an urban terrain reconstruction module.

[0040] The urban terrain point cloud processing module includes an urban terrain point cloud acquisition unit, an urban terrain feature object standard point cloud storage unit, a target urban terrain feature object standard point cloud search unit, an urban terrain semantic point cloud generation unit, an urban terrain semantic point cloud fusion unit, and an urban terrain semantic point cloud fusion preprocessing unit.

[0041] The urban terrain point cloud acquisition unit acquires urban terrain point cloud data using a vehicle-mounted LiDAR; the urban terrain feature object standard point cloud storage unit stores a database of urban terrain feature objects; the target urban terrain feature object standard point cloud search unit searches the target urban terrain feature object standard point cloud database based on the urban terrain point cloud data and in conjunction with a point cloud 3D modeling platform and the urban terrain feature object standard point cloud database; the urban terrain semantic point cloud generation unit generates the target urban terrain semantic point cloud database based on the urban terrain point cloud data and in conjunction with a point cloud 3D modeling platform and the target urban terrain feature object standard point cloud database. The system performs feature object attribute identification processing on the urban terrain point cloud using a standard point cloud database to obtain urban terrain semantic point cloud data. The urban terrain semantic point cloud fusion unit, based on the urban terrain semantic point cloud data and in conjunction with a point cloud 3D modeling platform, performs fusion processing on the urban terrain semantic point cloud and the standard point cloud of the target urban terrain feature object using the standard point cloud database to obtain urban terrain semantic point cloud fused data. The urban terrain semantic point cloud fusion preprocessing unit, based on the urban terrain semantic point cloud fusion data and in conjunction with a point cloud 3D modeling platform, performs noise reduction preprocessing on the urban terrain semantic point cloud fusion data to obtain urban terrain semantic point cloud fusion preprocessed data.

[0042] The urban terrain modeling and defect analysis module includes an urban terrain 3D model modeling unit, an urban terrain 3D model image acquisition unit, an urban terrain feature object standard modeling defect image storage unit, and an urban terrain 3D model modeling defect identification unit.

[0043] The urban terrain 3D modeling unit, based on the urban terrain semantic point cloud fusion preprocessing dataset, combines with the point cloud 3D modeling platform to perform 3D modeling processing of feature objects in the target urban terrain, obtaining urban terrain 3D model data; the urban terrain 3D model image acquisition unit acquires urban terrain 3D model image data through the point cloud 3D modeling platform and screenshot software; the urban terrain feature object standard modeling defect image storage unit is used to store urban terrain feature object standard modeling defect image data; the urban terrain 3D model modeling defect identification unit, based on the urban terrain 3D model image data and combined with AI algorithms and urban terrain feature object standard modeling defect image data, performs urban terrain 3D model modeling defect identification processing, obtaining urban terrain 3D model modeling defect identification information;

[0044] The urban terrain reconstruction module includes an urban terrain 3D model repair unit and an urban terrain 3D model output unit.

[0045] The urban terrain 3D model repair unit repairs and reconstructs the urban terrain 3D model based on the urban terrain 3D model data and in conjunction with the point cloud 3D modeling platform to obtain urban terrain 3D reconstruction model data; the urban terrain 3D model output unit outputs the urban terrain 3D model based on the urban terrain 3D reconstruction model data or the urban terrain 3D model data and in conjunction with the point cloud 3D modeling platform and the display screen.

[0046] This invention provides a method and system for rapid urban 3D terrain modeling based on AI point cloud semantic segmentation. It has the following beneficial effects:

[0047] I. Dynamically acquire urban terrain point cloud data using vehicle-mounted LiDAR, providing reliable data support for accurate urban terrain modeling based on semantic segmentation; accurately search the standard point cloud database of feature objects in the target urban terrain using urban terrain point cloud data and a point cloud 3D modeling platform, achieving standardized evaluation of urban terrain feature objects and improving the accuracy of modeling with limited point cloud data; autonomously identify the attributes of urban terrain point cloud feature objects based on urban terrain point cloud data, using a point cloud 3D modeling platform and a standard point cloud database of target urban terrain feature objects, realizing semantic segmentation of urban terrain. Point cloud classification and labeling improves the quality of urban terrain modeling. Based on urban terrain semantic point cloud data and in conjunction with a point cloud 3D modeling platform and a standard point cloud database of target urban terrain feature objects, precise fusion of the semantic point cloud of urban terrain with the standard point cloud of urban terrain feature objects is achieved. This ensures accurate supplementation of the spatial feature point cloud data required for urban terrain modeling, improving the accuracy of modeling with limited point cloud data. Furthermore, based on the fused semantic point cloud data of urban terrain and in conjunction with a point cloud 3D modeling platform, scientific noise reduction preprocessing of the fused semantic point cloud data of urban terrain is performed, achieving efficient filtering of interfering point cloud data for urban terrain modeling and improving the scientific rigor of modeling with limited point cloud data.

[0048] Second, based on the semantic point cloud fusion preprocessing dataset of urban terrain, and in conjunction with the point cloud 3D modeling platform, high-quality 3D modeling of feature objects in the target urban terrain is carried out. This achieves accurate modeling of urban terrain under limited point cloud data conditions based on semantic segmentation and point cloud fusion. By accurately collecting image information of urban terrain 3D model through the point cloud 3D modeling platform and screenshot software, and combining AI algorithms with standard modeling defect image data of urban terrain feature objects, intelligent identification of modeling defects in urban terrain 3D model is carried out. This achieves closed-loop self-checking of modeling results under limited point cloud data conditions of urban terrain, and improves the intelligence of urban terrain 3D model modeling.

[0049] Third, based on urban terrain 3D model data and in conjunction with a point cloud 3D modeling platform, dynamic and efficient repair and reconstruction processing of defects in urban terrain 3D modeling is carried out. This enables accurate modeling, repair and reconstruction of urban terrain under limited point cloud data conditions based on model repair, improving the reliability of urban terrain 3D modeling. Based on urban terrain 3D reconstruction model data or urban terrain 3D model data in conjunction with a point cloud 3D modeling platform and display screen, timely and safe output feedback of urban terrain 3D model is provided, improving the intuitiveness of feedback on urban terrain 3D modeling results. Attached Figure Description

[0050] Figure 1 A schematic diagram of the modules of the rapid urban 3D terrain modeling system based on AI point cloud semantic segmentation provided by the present invention;

[0051] Figure 2 The flowchart illustrates the rapid urban 3D terrain modeling method based on AI point cloud semantic segmentation provided by this invention. Detailed Implementation

[0052] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0053] An example of the rapid urban 3D terrain modeling method and system based on AI point cloud semantic segmentation is as follows:

[0054] Example 1: Please refer to Figures 1-2 A rapid urban 3D terrain modeling method based on AI point cloud semantic segmentation, the method includes the following steps:

[0055] The process involves collecting urban terrain point cloud data, searching a standard point cloud database of feature objects in the target urban terrain to obtain a standard point cloud database of feature objects, identifying the attributes of feature objects in the urban terrain point cloud to obtain semantic point cloud data of the urban terrain, fusing the semantic point cloud of the urban terrain with the standard point cloud of the urban terrain feature objects to obtain fused semantic point cloud data of the urban terrain, performing data denoising preprocessing on the fused semantic point cloud data of the urban terrain to obtain fused preprocessed semantic point cloud data of the urban terrain, modeling the 3D model of feature objects in the target urban terrain to obtain 3D model data of the urban terrain, collecting image data of the 3D model of the urban terrain, identifying modeling defects in the 3D model of the urban terrain to obtain modeling defect identification information of the 3D model of the urban terrain, directly outputting the 3D model of the urban terrain when there are no defects, and repairing and reconstructing the modeling defects of the 3D model of the urban terrain when defects exist to obtain 3D reconstructed model data of the urban terrain and outputting it.

[0056] For further details, please refer to Figures 1-2 The process involves collecting urban terrain point cloud data, searching and processing the standard point cloud database of feature objects in the target urban terrain to obtain a standard point cloud database of target urban terrain feature objects, identifying the attributes of the feature objects in the urban terrain point cloud to obtain urban terrain semantic point cloud data, fusing the urban terrain semantic point cloud with the standard point cloud of urban terrain feature objects to obtain urban terrain semantic point cloud fused data, and performing data noise reduction preprocessing on the urban terrain semantic point cloud fused data to obtain urban terrain semantic point cloud fused preprocessed data. The steps are as follows:

[0057] Step 11: Collect and process point cloud data of feature objects in the surrounding terrain environment of the autonomous vehicle using vehicle-mounted LiDAR to obtain a set of urban terrain point cloud data. ,in Indicates the number of collections Urban terrain point cloud data; the urban terrain point cloud data includes the location and color information of the outline surface of the target city's terrain feature objects;

[0058] Step 12: Based on the urban terrain point cloud data set and the standard point cloud database matrix of urban terrain feature objects, perform search processing on the standard point cloud database of feature objects in the target urban terrain to obtain the standard point cloud database set of target urban terrain feature objects; perform feature object attribute identification processing on the urban terrain point cloud according to the urban terrain point cloud data set and the standard point cloud database set of target urban terrain feature objects to obtain the urban terrain semantic point cloud data set.

[0059] Step 13: Based on the urban terrain semantic point cloud data set and the target urban terrain feature object standard point cloud database set, perform fusion processing of the urban terrain semantic point cloud and the urban terrain feature object standard point cloud to obtain the urban terrain semantic point cloud fusion data set; perform noise reduction preprocessing of the urban terrain semantic point cloud fusion data set based on the urban terrain semantic point cloud fusion data set to obtain the urban terrain semantic point cloud fusion preprocessed data set.

[0060] The following steps are taken to search the standard point cloud database of feature objects in the target city's terrain based on the urban terrain point cloud dataset and the standard point cloud database matrix of urban terrain feature objects, to obtain the target city terrain feature object standard point cloud database set. The steps for performing feature object attribute identification processing on the urban terrain point cloud dataset and the target city terrain feature object standard point cloud database set to obtain the urban terrain semantic point cloud dataset are as follows:

[0061] Step 121: Establish a standard point cloud database matrix for urban terrain feature objects. ,in Indicates the first The standard point cloud database of urban terrain feature objects corresponds to various types of urban terrain feature objects. These types include different types of ground, different types of buildings, different types of vehicles, different types of transportation equipment, different types of vegetation, and different types of people. The standard point cloud database of urban terrain feature objects represents a collection of point cloud data of the complete spatial contour surface of the feature objects set for different types of urban terrain feature objects. ,in Standard point cloud database representing urban terrain features The Middle Standard point cloud data for urban terrain features, including location information, color information, and type attribute information of the outline surface of urban terrain features; wherein the type attribute information of urban terrain features includes any one or more of the following: name information, specification information, type information, gender information, and material information of urban terrain features.

[0062] Step 122: Use a point cloud 3D modeling platform to collect urban terrain point cloud data. All city terrain point cloud data The spatial point cloud contour features of urban terrain features and the standard point cloud database matrix of urban terrain features. Internal urban terrain feature object standard point cloud database Standard point cloud data of all urban terrain features Spatial point cloud contour features of standard urban terrain feature objects are matched to search for matching urban terrain point cloud datasets. A standard point cloud database for matching urban terrain feature objects with spatial point cloud contour features. And through data identification, a standard point cloud database set of target city terrain feature objects is generated. , ,in and They represent the first species and first A standard point cloud database of target city terrain feature objects corresponding to various types of city terrain feature objects. ,in and These represent the standard point cloud databases of the target city's terrain features. The Middle The and the first Standard point cloud data of the terrain features of each target city; ,in and These represent the standard point cloud databases of the target city's terrain features. The Middle The and the first Standard point cloud data of target city terrain feature objects, the point cloud 3D modeling platform includes any one of Geomagic Studio, CloudCompare, and Open3D; the standard point cloud data of target city terrain feature objects includes the position information, color information and type attribute information of the outline surface of the target city terrain feature objects.

[0063] Step 123: Based on the point cloud 3D modeling platform, combine the urban terrain point cloud data set. Urban topographic point cloud data A collection of standard point cloud databases for matching spatial point cloud contour features of urban terrain feature objects. Standard point cloud database of target city terrain features to The corresponding urban terrain feature object type text information is mapped to urban terrain point cloud data that matches the spatial point cloud contour features of the target urban terrain feature object in the standard point cloud database. And through data identification, a set of urban terrain semantic point cloud data is generated. ,in Indicates the first The city terrain semantic point cloud data includes the location and color information of the outline surface of the target city terrain feature object, as well as the type attribute information of the target city terrain feature object.

[0064] The steps for fusing the semantic point cloud of urban terrain with the standard point cloud database of target urban terrain feature objects are as follows: First, a semantic point cloud of urban terrain is fused with the standard point cloud of urban terrain feature objects to obtain a fused semantic point cloud dataset. Second, a denoising preprocessing step is performed on the fused semantic point cloud dataset to obtain a preprocessed semantic point cloud dataset.

[0065] Step 131: Based on the point cloud 3D modeling platform, combine it with the urban terrain point cloud data set. Urban topographic point cloud data A collection of standard point cloud databases for matching spatial point cloud contour features of urban terrain feature objects. Internal target city terrain feature object standard point cloud database to Set of standard point cloud data and semantic point cloud data of urban terrain features in the target city semantic point cloud data of urban terrain Point cloud data is combined by matching the spatial point cloud contour features of urban terrain features, and a semantic point cloud fusion dataset of urban terrain is generated. , > ,in Indicates the first Urban terrain semantic point cloud fusion data; the urban terrain semantic point cloud fusion data includes the position and color information of the fused outline surface of the target urban terrain feature object, the type attribute information of the target urban terrain feature object, and the position and color information of the target urban terrain feature object after fusion;

[0066] Step 132: Based on the point cloud 3D modeling platform, fuse the urban terrain semantic point cloud data set. Located in the middle, it is composed of semantic point cloud fusion data of all urban terrain. The urban terrain feature object spatial point cloud features are composed of urban terrain semantic point cloud fusion data. Perform data filtering and preprocessing, and then combine the data from the fusion of semantic point clouds of all urban terrain features. The urban terrain features are formed by the spatial point cloud feature contour surface of the urban terrain semantic point cloud fusion data. Data identification is performed to generate a preprocessed dataset for urban terrain semantic point cloud fusion. , ≤ ,in Indicates the first The preprocessed data for urban terrain semantic point cloud fusion includes the location and color information of the fusion contour surface of the target urban terrain feature object, and the type attribute information of the target urban terrain feature object.

[0067] By dynamically acquiring urban terrain point cloud data using vehicle-mounted LiDAR, reliable data support is provided for accurate urban terrain modeling based on semantic segmentation. Based on the urban terrain point cloud data and a standard point cloud database of urban terrain feature objects, and in conjunction with a point cloud 3D modeling platform, accurate searches of the standard point cloud database of feature objects in the target urban terrain are performed, achieving standardized evaluation of urban terrain feature objects and improving the accuracy of modeling with limited point cloud data. Furthermore, based on the urban terrain point cloud data and in conjunction with the point cloud 3D modeling platform and the standard point cloud database of target urban terrain feature objects, the attributes of urban terrain point cloud feature objects are autonomously identified, enabling semantic point cloud modeling of urban terrain based on semantic segmentation. Cloud classification and labeling processing improves the quality of urban terrain modeling; based on urban terrain semantic point cloud data and in conjunction with the point cloud 3D modeling platform and the standard point cloud database of the target urban terrain feature objects, the semantic point cloud of urban terrain is accurately fused with the standard point cloud of urban terrain feature objects, realizing the accurate supplementation of spatial feature point cloud data required for urban terrain modeling and improving the accuracy of urban terrain modeling with limited point cloud data; based on the urban terrain semantic point cloud fusion data and in conjunction with the point cloud 3D modeling platform, the semantic point cloud fusion data of urban terrain is scientifically denoised and preprocessed to achieve efficient filtering of interfering point cloud data for urban terrain modeling, improving the scientific nature of urban terrain modeling with limited point cloud data.

[0068] For further details, please refer to Figures 1-2 The process involves modeling 3D models of feature objects in the target city's terrain to obtain 3D urban terrain model data; collecting 3D urban terrain model image data; identifying and processing modeling defects in the 3D urban terrain model to obtain 3D urban terrain modeling defect identification information; and directly outputting the 3D urban terrain model when no defects are found. The steps are as follows:

[0069] Step 21: Based on the urban terrain semantic point cloud fusion preprocessing data set, perform 3D modeling of feature objects in the target urban terrain to obtain urban terrain 3D model data;

[0070] Step 22: Based on the urban terrain 3D model data, perform appearance image acquisition and processing of the urban terrain 3D model to obtain an urban terrain 3D model image data set. Based on the urban terrain 3D model image data set and the urban terrain feature object standard modeling defect image data matrix, perform urban terrain 3D model modeling defect identification processing to obtain urban terrain 3D model modeling defect identification information. The urban terrain 3D model modeling defect identification information includes no defects and defects. When there are no defects, the urban terrain 3D model is directly output.

[0071] The steps for performing 3D modeling of feature objects in the target city's terrain based on the preprocessed urban terrain semantic point cloud fusion dataset to obtain the 3D urban terrain model data are as follows:

[0072] Step 211: Obtain the preprocessed data set of urban terrain semantic point cloud fusion. ;

[0073] Step 212: Preprocess the data set based on the semantic point cloud fusion of urban terrain using a point cloud 3D modeling platform. Preprocessed data of urban terrain semantic point cloud fusion The corresponding target city terrain feature object integrates the position and color information of the outline surface and the type attribute information of the target city terrain feature object to perform three-dimensional modeling processing of the feature object in the target city terrain, and generates urban terrain three-dimensional model data, which represents the surface outline three-dimensional model information of the target city terrain feature object.

[0074] Based on the urban terrain 3D model data, the appearance images of the urban terrain 3D model are acquired and processed to obtain an urban terrain 3D model image dataset. Based on the urban terrain 3D model image dataset and the standard modeling defect image data matrix of urban terrain feature objects, modeling defect identification processing is performed on the urban terrain 3D model to obtain urban terrain 3D model modeling defect identification information. This information includes whether defects are present or absent. When no defects are present, the operation steps for directly outputting the urban terrain 3D model are as follows:

[0075] Step 221: Input the 3D urban terrain model data into the point cloud 3D modeling platform and run it. Simultaneously, use screenshot software to capture screenshots of the displayed 3D urban terrain model and generate a set of 3D urban terrain model image data. ,in Indicates the number of collections Three-dimensional model image data of urban terrain;

[0076] Step 222: Establish a standard modeling defect image data matrix for urban terrain feature objects. ,in Indicates the first The image data represents the standard modeling defects of urban terrain feature objects. The standard modeling defects of urban terrain feature objects represent the appearance image information of the standard urban terrain 3D model set for the defect types of point cloud 3D modeling of urban terrain feature objects. The defect types of point cloud 3D modeling of urban terrain feature objects include gaps on the model surface, uneven model surface, irregular bumps on the model surface, burrs or isolated floating patches on the model surface, and sharp edges on the model surface.

[0077] Step 223: Collect the 3D urban terrain model image data. Three-dimensional model image data of urban terrain Standard modeling defect image data matrix of urban terrain feature objects Standard modeling defect image data of urban terrain feature objects Image feature matching is performed, and based on the matching results, defect identification information for the 3D urban terrain model is generated. The specific steps for generating this defect identification information are as follows:

[0078] Step 2231, Input Image Preprocessing: Preprocess the urban terrain 3D model image data set. Three-dimensional model image data of urban terrain and the standard modeling defects of urban terrain feature objects image data matrix Standard modeling defect image data of urban terrain feature objects The image data is preprocessed by scaling the image to a fixed size and normalizing the image.

[0079] Step 2232, Convolutional Layer Processing: Processing the preprocessed 3D urban terrain model image data and urban terrain feature object standard modeling defect image data Convolution is used for processing; the convolution formula is: ,in and These represent the first convolution adjustment factor and the second convolution adjustment factor, respectively. Represents a constant. The input variable for the convolution formula is the preprocessed 3D urban terrain model image data. and urban terrain feature object standard modeling defect image data ; express The convolution objective function, The base is The sum of the exponents is The exponential function; Represents the convolution objective function and exponential function In the interval The convolution value, Indicates the input variables of the convolution formula Differential; extracting convolutional data to form a new 3D urban terrain model image. and urban terrain feature object standard modeling defect image data ;

[0080] Step 2233: Activation function layer processing: Process the 3D urban terrain model image data output by the convolutional layer. and urban terrain feature object standard modeling defect image data Nonlinear transformations are performed using the ReLU activation function, the formula for which is: ,in Represents 3D model image data of urban terrain and urban terrain feature object standard modeling defect image data The convolution value; Indicates the value Random numbers within the interval;

[0081] Steps 2234: Pooling layer processing: Processing the linearly transformed 3D urban terrain model image data... and urban terrain feature object standard modeling defect image data Perform a downsampling operation;

[0082] Step 2235, Fully Connected Layer Processing: Process the downsampled 3D urban terrain model image data and urban terrain feature object standard modeling defect image data Flattened into vectors, these vectors can be connected to fully connected layers for classification or regression.

[0083] Step 2236, Output Results: Output the 3D urban terrain model image data for the fully connected layer. and urban terrain feature object standard modeling defect image data Image matching processing is performed based on the 3D urban terrain model image data. and urban terrain feature object standard modeling defect image data Image feature matching results are used to generate a 3D urban terrain model and identify defects in the modeling process.

[0084] when and If image feature matching is successful, it indicates that there are defects in the modeling of urban terrain features and the model needs to be repaired and rebuilt. The output of the urban terrain 3D model modeling defect identification information is: there are defects.

[0085] when and If no matching of image features is found, it indicates that there are no defects in the modeling of urban terrain features. In this case, the output of the urban terrain 3D model modeling defect identification information is that there are no defects. At this time, the urban terrain 3D model data is directly input into the point cloud 3D modeling platform and run in conjunction with the display screen to output the urban terrain 3D model.

[0086] Based on the semantic point cloud fusion preprocessing dataset of urban terrain, and in conjunction with the point cloud 3D modeling platform, high-quality 3D modeling of feature objects in the target urban terrain is achieved. This enables accurate modeling of urban terrain under limited point cloud data conditions based on semantic segmentation and point cloud fusion. By accurately collecting image information of urban terrain 3D models through the point cloud 3D modeling platform and screenshot software, and combining AI algorithms with standard modeling defect image data of urban terrain feature objects, intelligent identification of modeling defects in urban terrain 3D models is performed. This achieves closed-loop self-checking of modeling results under limited point cloud data conditions, improving the intelligence of urban terrain 3D modeling.

[0087] For further details, please refer to Figures 1-2 When defects exist, the following steps are taken to repair and reconstruct the 3D urban terrain model to obtain and output the 3D urban terrain reconstruction model data:

[0088] Step 31: When the defect identification information of the urban terrain 3D model is that there is a defect, the urban terrain 3D model data in the urban terrain 3D model data is used to perform defect detection, repair and reconstruction of the model outline surface modeling of the urban terrain 3D model through the point cloud 3D modeling platform, and generate urban terrain 3D reconstruction model data.

[0089] Step 32: Input the 3D reconstruction model data of the urban terrain into the point cloud 3D modeling platform, run it, and output the 3D model of the urban terrain with the help of the display screen.

[0090] Based on urban terrain 3D model data and in conjunction with a point cloud 3D modeling platform, dynamic and efficient repair and reconstruction processing of defects in urban terrain 3D modeling is performed. This enables accurate modeling, repair, and reconstruction of urban terrain under limited point cloud data conditions, improving the reliability of urban terrain 3D modeling. Based on urban terrain 3D reconstruction model data or urban terrain 3D model data in conjunction with a point cloud 3D modeling platform and display screen, timely and secure output feedback of urban terrain 3D model is provided, improving the intuitiveness of the feedback of urban terrain 3D modeling results.

[0091] Example 2: Please refer to Figures 1-2A rapid urban 3D terrain modeling system based on AI point cloud semantic segmentation is used to realize a rapid urban 3D terrain modeling method based on AI point cloud semantic segmentation. The system includes an urban terrain point cloud processing module, an urban terrain modeling and defect analysis module, and an urban terrain reconstruction module.

[0092] The urban terrain point cloud processing module includes an urban terrain point cloud acquisition unit, an urban terrain feature object standard point cloud storage unit, a target urban terrain feature object standard point cloud search unit, an urban terrain semantic point cloud generation unit, an urban terrain semantic point cloud fusion unit, and an urban terrain semantic point cloud fusion preprocessing unit.

[0093] The system comprises the following components: an urban terrain point cloud acquisition unit, which acquires urban terrain point cloud data via a vehicle-mounted LiDAR; an urban terrain feature object standard point cloud storage unit, which stores a database of urban terrain feature objects standard point clouds; a target urban terrain feature object standard point cloud search unit, which searches for standard point clouds of feature objects in the target urban terrain based on the urban terrain point cloud data and in conjunction with a point cloud 3D modeling platform and the urban terrain feature object standard point cloud database; and an urban terrain semantic point cloud generation unit, which generates semantic point clouds based on the urban terrain point cloud data and in conjunction with a point cloud 3D modeling platform and the target urban terrain feature object standard point cloud database. The quasi-point cloud database performs feature object attribute identification processing on the urban terrain point cloud to obtain urban terrain semantic point cloud data; the urban terrain semantic point cloud fusion unit, based on the urban terrain semantic point cloud data and in conjunction with the point cloud 3D modeling platform and the target urban terrain feature object standard point cloud database, performs fusion processing on the urban terrain semantic point cloud and the standard point cloud of the urban terrain feature object to obtain urban terrain semantic point cloud fused data; the urban terrain semantic point cloud fusion preprocessing unit, based on the urban terrain semantic point cloud fusion data and in conjunction with the point cloud 3D modeling platform, performs noise reduction preprocessing on the urban terrain semantic point cloud fusion data to obtain urban terrain semantic point cloud fusion preprocessed data.

[0094] The urban terrain modeling and defect analysis module includes an urban terrain 3D model modeling unit, an urban terrain 3D model image acquisition unit, an urban terrain feature object standard modeling defect image storage unit, and an urban terrain 3D model modeling defect identification unit.

[0095] The system comprises the following components: a city terrain 3D modeling unit, which merges a preprocessed dataset of city terrain semantic point cloud with a point cloud 3D modeling platform to perform 3D modeling of feature objects in the target city terrain, resulting in city terrain 3D model data; a city terrain 3D model image acquisition unit, which acquires city terrain 3D model image data through a point cloud 3D modeling platform and screenshot software; a city terrain feature object standard modeling defect image storage unit, which stores standard modeling defect image data of city terrain feature objects; and a city terrain 3D model modeling defect identification unit, which, based on the city terrain 3D model image data and combined with AI algorithms and standard modeling defect image data of city terrain feature objects, performs modeling defect identification processing on the city terrain 3D model to obtain city terrain 3D model modeling defect identification information.

[0096] The urban terrain reconstruction module includes an urban terrain 3D model repair unit and an urban terrain 3D model output unit.

[0097] The urban terrain 3D model repair unit repairs and reconstructs the urban terrain 3D model based on the urban terrain 3D model data and in conjunction with the point cloud 3D modeling platform to obtain urban terrain 3D reconstruction model data; the urban terrain 3D model output unit outputs the urban terrain 3D model based on the urban terrain 3D reconstruction model data or the urban terrain 3D model data and in conjunction with the point cloud 3D modeling platform and the display screen.

[0098] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A rapid urban 3D terrain modeling method based on AI point cloud semantic segmentation, characterized in that, The method includes the following steps: The process involves collecting urban terrain point cloud data, searching and processing a standard point cloud database of feature objects in the target urban terrain to obtain a standard point cloud database of target urban terrain feature objects, identifying the attributes of the feature objects in the urban terrain point cloud to obtain urban terrain semantic point cloud data, fusing the semantic point cloud of the urban terrain with the standard point cloud of the urban terrain feature objects to obtain fused urban terrain semantic point cloud data, and performing data denoising preprocessing on the fused urban terrain semantic point cloud data to obtain fused urban terrain semantic point cloud data. The preprocessed urban terrain semantic point cloud data includes the following steps: By collecting and processing point cloud data of feature objects in the surrounding terrain environment of an autonomous vehicle using an onboard LiDAR, a set of urban terrain point cloud data is obtained. The include ;in Indicates the number of collections Urban terrain point cloud data, the urban terrain point cloud data including the location information and color information of the outline surface of the target urban terrain feature object; Based on the urban terrain point cloud data set and the standard point cloud database matrix of urban terrain feature objects, a search process is performed on the standard point cloud database of feature objects in the target urban terrain to obtain a standard point cloud database set of target urban terrain feature objects; based on the urban terrain point cloud data set and the standard point cloud database set of target urban terrain feature objects, feature object attribute identification processing is performed on the urban terrain point cloud to obtain an urban terrain semantic point cloud data set. Based on the urban terrain semantic point cloud data set and the target urban terrain feature object standard point cloud database set, the semantic point cloud of urban terrain and the standard point cloud of urban terrain feature object are fused to obtain an urban terrain semantic point cloud fused data set; based on the urban terrain semantic point cloud fused data set, the urban terrain semantic point cloud fused data is denoised and preprocessed to obtain an urban terrain semantic point cloud fused preprocessed data set. The process involves: modeling a 3D model of a feature object in the target city's terrain to obtain 3D urban terrain model data; collecting 3D urban terrain model image data; identifying and processing modeling defects in the 3D urban terrain model to obtain 3D urban terrain modeling defect identification information; and directly outputting the 3D urban terrain model when no defects are found. The steps include: Based on the urban terrain semantic point cloud fusion preprocessing data set, the three-dimensional modeling process of feature objects in the target urban terrain is carried out to obtain urban terrain three-dimensional model data. Based on the urban terrain 3D model data, the appearance image of the urban terrain 3D model is acquired and processed to obtain an urban terrain 3D model image data set. Based on the urban terrain 3D model image data set and the standard modeling defect image data matrix of urban terrain feature objects, modeling defect identification processing of the urban terrain 3D model is performed to obtain urban terrain 3D model modeling defect identification information. The urban terrain 3D model modeling defect identification information includes no defects and defects present. When no defects are present, the urban terrain 3D model is directly output, including the following steps: The urban terrain 3D model data is input into a point cloud 3D modeling platform and run, and displayed in all directions on a screen. Simultaneously, screenshot software is used to capture images of the displayed urban terrain 3D model, generating a set of urban terrain 3D model image data. The include ;in Indicates the number of collections Three-dimensional model image data of urban terrain; Establish a standard modeling defect image data matrix for urban terrain feature objects The include ;in Indicates the first Image data with defects in standard modeling of urban terrain features; The The above With the The above Image feature matching is performed, and based on the image feature matching results, defect identification information for the 3D urban terrain model is generated. The specific steps for generating the defect identification information for the 3D urban terrain model are as follows: Step 2231, Input Image Preprocessing: [The text appears to be incomplete and contains several grammatical errors. A more accurate translation would require the full context.] The above and stated The above The image data is preprocessed by scaling the image to a fixed size and normalizing the image. Step 2232, Convolutional Layer Processing: Process the pre-processed... and stated Convolution processing is employed; after extracting the convolution, a new [process] is formed. and stated ; Step 2233, Activation function layer processing: Process the output of the convolutional layer as described above. and stated Nonlinear transformation is performed using the ReLU activation function; Step 2234, Pooling layer processing: For the linear transformation described above... and stated Perform a downsampling operation; Step 2235, Fully Connected Layer Processing: Process the downsampled layer... and stated Flattened into vectors, these vectors can be connected to fully connected layers for classification or regression. Step 2236, Output Results: The output of the fully connected layer is as follows. and stated Perform image matching processing, based on the above. and stated Image feature matching results are used to generate a 3D urban terrain model and identify defects in the modeling process. when and If image feature matching is successful, the defect identification information of the three-dimensional urban terrain model is output as "defect exists". when and If no matching of image features is found, the defect identification information of the urban terrain 3D model is output as no defect exists. At this time, the urban terrain 3D model data is directly input into the point cloud 3D modeling platform and the urban terrain 3D model is output in conjunction with the display screen. When defects exist, the modeling defects of the urban terrain 3D model are repaired and reconstructed to obtain urban terrain 3D reconstruction model data and output it.

2. The rapid urban 3D terrain modeling method based on AI point cloud semantic segmentation according to claim 1, characterized in that: The following steps are taken to perform a search on the standard point cloud database of feature objects in the target city terrain based on the urban terrain point cloud dataset and the standard point cloud database matrix of urban terrain feature objects, to obtain a standard point cloud database set of target city terrain feature objects. The steps for performing feature object attribute identification processing on the urban terrain point cloud dataset and the standard point cloud database set of target city terrain feature objects to obtain the urban terrain semantic point cloud dataset are as follows: Establish a standard point cloud database matrix for urban terrain feature objects. The include ;in Indicates the first The standard point cloud database of urban terrain feature objects corresponding to the various types of urban terrain feature objects; include ;in Indicates the The Middle Standard point cloud data of urban terrain features; The point cloud 3D modeling platform is used to... All of the above The urban terrain features formed by the spatial point cloud contour features and the aforementioned Internal description All of the above The spatial point cloud contour features of the standard urban terrain feature objects are matched with the spatial point cloud contour features to search for matching elements. The matching of spatial point cloud contour features of urban terrain feature objects And through data identification, a standard point cloud database set of target city terrain feature objects is generated. The include and ,in and They represent the first species and first The standard point cloud database of target city terrain feature objects corresponding to the various types of city terrain feature objects, the include and ,in and Each represents the The Middle The and the first Standard point cloud data of target city terrain features; include and ,in and Each represents the The Middle The and the first Standard point cloud data of target city terrain feature objects, including the location information, color information and type attribute information of the outline surface of the target city terrain feature objects; Based on the point cloud 3D modeling platform, it will be combined with the aforementioned The above The matching of spatial point cloud contour features of urban terrain feature objects The above to The corresponding urban terrain feature object type text information is mapped to the standard point cloud database of the target urban terrain feature object to match the spatial point cloud contour features of the urban terrain feature object. And through data identification, a set of urban terrain semantic point cloud data is generated. The include ;in Indicates the first The city terrain semantic point cloud data includes the location and color information of the outline surface of the target city terrain feature object, as well as the type attribute information of the target city terrain feature object.

3. The rapid urban 3D terrain modeling method based on AI point cloud semantic segmentation according to claim 2, characterized in that: The steps for fusing the semantic point cloud of urban terrain with the standard point cloud database of the target urban terrain feature objects based on the urban terrain semantic point cloud dataset are as follows: First, the semantic point cloud of urban terrain is fused with the standard point cloud of the urban terrain feature objects to obtain a fused semantic point cloud dataset. Second, the semantic point cloud fusion dataset of urban terrain is denoised based on the fused semantic point cloud dataset to obtain a preprocessed semantic point cloud fusion dataset. The point cloud-based 3D modeling platform will be combined with the aforementioned The above The matching of spatial point cloud contour features of urban terrain feature objects Internal description to The standard point cloud data of the target city terrain feature object described in the text and the... The above Point cloud data is combined by matching the spatial point cloud contour features of urban terrain features, and a semantic point cloud fusion dataset of urban terrain is generated. The include ;in Indicates the first The urban terrain semantic point cloud fusion data includes the position and color information of the fused outline surface of the target urban terrain feature object, the type attribute information of the target urban terrain feature object, and the position and color information of the target urban terrain feature object after fusion. Based on the point cloud 3D modeling platform, the aforementioned Located in the middle, by all the described The urban terrain features formed by the spatial point cloud features within the object Perform data filtering preprocessing, and locate the data in the format described by all of the above. The urban terrain feature object spatial point cloud feature contour surface constituted by the Data identification is performed to generate a preprocessed dataset for urban terrain semantic point cloud fusion. The include ,in Indicates the first The preprocessed data for the fusion of semantic point clouds of urban terrain includes the position and color information of the fused contour surface of the target urban terrain feature object, and the type attribute information of the target urban terrain feature object.

4. The rapid urban 3D terrain modeling method based on AI point cloud semantic segmentation according to claim 3, characterized in that: The steps for performing 3D modeling of feature objects in the target city terrain based on the aforementioned urban terrain semantic point cloud fusion preprocessing data set to obtain urban terrain 3D model data are as follows: Obtain the ; Based on the point cloud 3D modeling platform The above The corresponding target city terrain feature object integrates the location and color information of the outline surface and the type attribute information of the target city terrain feature object to perform three-dimensional modeling of the feature object in the target city terrain, and generates three-dimensional model data of the city terrain.

5. The rapid urban 3D terrain modeling method based on AI point cloud semantic segmentation according to claim 4, characterized in that: When defects exist, the steps to repair and reconstruct the urban terrain 3D model to obtain and output the urban terrain 3D reconstruction model data are as follows: When the defect identification information of the urban terrain 3D model is that there is a defect, the urban terrain 3D model data in the urban terrain 3D model data is used to perform defect detection, repair and reconstruction work on the model outline surface modeling of the urban terrain 3D model through the point cloud 3D modeling platform, and urban terrain 3D reconstruction model data is generated. The urban terrain 3D reconstruction model data is input into the point cloud 3D modeling platform and run, and the urban terrain 3D model is output in conjunction with the display screen.

6. A rapid urban 3D terrain modeling system based on AI point cloud semantic segmentation, used to implement the rapid urban 3D terrain modeling method based on AI point cloud semantic segmentation as described in any one of claims 1-5, characterized in that: The system includes an urban terrain point cloud processing module, an urban terrain modeling and defect analysis module, and an urban terrain reconstruction module. The urban terrain point cloud processing module is used to collect urban terrain point cloud data, search and process the standard point cloud database of feature objects in the target urban terrain to obtain the standard point cloud database of feature objects of the target urban terrain; identify the attributes of feature objects in the urban terrain point cloud to obtain urban terrain semantic point cloud data; fuse the semantic point cloud of the urban terrain with the standard point cloud of urban terrain feature objects to obtain urban terrain semantic point cloud fused data; and perform data noise reduction preprocessing on the semantic point cloud fused data of the urban terrain to obtain urban terrain semantic point cloud fused preprocessed data. The urban terrain modeling and defect analysis module is used to model the three-dimensional model of the feature objects in the target urban terrain to obtain urban terrain three-dimensional model data; collect urban terrain three-dimensional model image data; identify and process the modeling defects of the urban terrain three-dimensional model to obtain urban terrain three-dimensional model modeling defect identification information. The urban terrain reconstruction module is used to repair and reconstruct the modeling defects of the urban terrain 3D model, obtain the urban terrain 3D reconstruction model data, and output it.

Images